Flexible adaptation

  • Koalas declared endangered in NSW, QLD and ACT

  • Climate change threatens koala habitat

  • NSW Koala Strategy: 7,000ha koala habitat to be protected on NSW private land by 2025

library(ggplot2)
## Warning: package 'ggplot2' was built under R version 4.1.3
library(patchwork)
## Warning: package 'patchwork' was built under R version 4.1.3
library(sf)
## Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 7.2.1; sf_use_s2() is TRUE
library(geodata)
## Warning: package 'geodata' was built under R version 4.1.3
## Loading required package: terra
## Warning: package 'terra' was built under R version 4.1.3
## terra 1.7.18
## 
## Attaching package: 'terra'
## The following object is masked from 'package:patchwork':
## 
##     area
plot_list <- readRDS('../plots/plot_list.rds')
shapes <- readRDS("../plots/shapes.rds")
aus_border <- shapes$aus_border
nsw_bbox <- shapes$nsw_bbox
nsw_lga_union <- shapes$nsw
bbox_buffer <- 1

wrap_elements(full=plot_list[[1]]$aus_plot)

7,000ha koala habitat on private land

Koala habitat: KITL index > 0.25

fcn_generate_plot <- function(i, plot_list) {
  layout <- "
  BBDDD
  CCDDD
  CCDDD"
  plot_list[[i]]$covenanted_area_plot + plot_list[[i]]$year_trend_plot + plot_list[[i]]$prop_decisions_plot + plot_layout(design = layout)
} 
fcn_generate_plot(1, plot_list)
CC: Enough protection under current conditions

CC: Enough protection under current conditions 

fcn_generate_plot(2, plot_list)
RI: Robust but inflexible protection

RI: Robust but inflexible protection 

fcn_generate_plot(3, plot_list)
F: Flexible protection -- add new covenants in 2050

F: Flexible protection – add new covenants in 2050 

fcn_generate_plot(4, plot_list)
F+L: Flexible protection + learning and adapting to climate change

F+L: Flexible protection + learning and adapting to climate change

Flexible adaptation halves conservation costs

plot_list[[1]]$cost_plot
CC: $70M ($13-138M)

CC: $70M ($13-138M) 

plot_list[[2]]$cost_plot
RI: $152M ($100-229M)

RI: $152M ($100-229M) 

plot_list[[3]]$cost_plot
F: $84M ($36-161M) -- 42% cost reduction

F: $84M ($36-161M) – 42% cost reduction 

plot_list[[4]]$cost_plot
Flexible + Learning: $78M ($25-166M) -- 44% cost reduction

Flexible + Learning: $78M ($25-166M) – 44% cost reduction

Flexibility: adding protection more valuable than ending

plot_list_ns <- readRDS('../plots/plot_list_ns.rds')
bar_width <- 0.2
plot_list_ns[[1]]
Flexibility to delay investments only

Flexibility to delay investments only 

plot_list_ns[[2]]
Flexibility to end interventions only

Flexibility to end interventions only 

plot_list_ns[[3]]
Flexbility to delay and end investments

Flexbility to delay and end investments

Key findings

  • Strategic flexibility alters first-stage decisions

  • Flexibility mitigates trade-offs between maximizing conservation outcomes and managing risks

  • Flexibility to offer new covenants much more valuable than flexibility to end interventions

Thank you to co-authors: Brooke Williams, Carla Archibald, James Brazill-Boast, Michael Drielsma and Jonathan Rhodes

Appendix 1: Sensitivity analysis

Sensitivities of estimates of the value of flexibility (A, E and A/E) under changes to the default parameters, showing sensitivities to a, sampling of 10 study populations randomly sampled with the stratified sampling approach (default index = 1), b, year covenant modification is allowed - t’ (default = 2050). c, koala landscape capacity indicator cut-off (default = 0.25) across a range of cut-offs where feasible solutions to the problem are found, and d. Amount of learning based on the number of climate scenarios decision-makers are uncertain about (1 being perfect certainty over climate change, default = 12), with “No Learning”, “Partial Learning” and “Full Learning” having parameters of 12, 3 and 1 respectively.

Appendix 2: Two-stage stochastic optimization

\[ \min_x E_{j \in J}\Big[\sum_i \sum_t c_{ijt}x_i + E_{S\in\xi}[Q(x,S)]\Big] \\ Q_j(x,S) := \min_{y,w} E_{s\in S}\Big[\sum_i \sum_{t\geq t'} c_{ijt}y_{is} - c_{ijt}w_{is}\Big] \\ s.t. \sum_i m_{ijts} x_i \geq K \quad \forall t=1,s\in S, j \in J \\ \sum_i m_{ijts}(x_i + y_{is} - w_{is} \geq K \quad \forall t \geq t', s \in S, j \in J \\ \sum_i m_{ijts} x_i \geq K \quad \forall t = 1, s \in S, j \in J \\ x_i + y_{is} \leq 1 \quad \forall i \in N, s \in S \\ x_i \geq w_{is} \quad \forall i \in N, s \in S \\ y_{is} = 0 \quad \forall i \in N, s \in S \\ w_{is} = 0 \quad \forall i \in N, s \in S \\ x,y,w \in [0,1] \]

Appendix 3

Conservation under Current Conditions (CC)

  • Total habitat reaches 7,000 ha in 2020 on average

Robust and Inflexible (RI)

  • Total habitat is higher than 7,000 for all time-steps, all climate scenarios and all social behavior realizations
  • Covenant offers same across time

Flexible (F)

  • Total habitat is higher than 7,000 for all time-steps, all climate scenarios and all social behavior realizations
  • New covenant offers can be added in 2050

Flexible with Learning (F+L)

  • Total habitat is higher than 7,000 for time-steps before 2050, all climate scenarios and all social behavior realizations
  • After 2050, total habitat must be higher than 7,000 for a particular climate scenario (offers can be different across scenarios) on the average social behavior realization
  • New covenant offers can be added in 2050